Abstract:Human-machine interfaces (HMI) play a pivotal role in the rehabilitation and daily assistance of lower-limb amputees. The brain of such interfaces is a control model that detects the user's intention using sensor input and generates corresponding output (control commands). With recent advances in technology, AI-based policies have gained attention as control models for HMIs. However, supervised learning techniques require affluent amounts of labeled training data from the user, which is challenging in the context of lower-limb rehabilitation. Moreover, a static pre-trained model does not take the temporal variations in the motion of the amputee (e.g., due to speed, terrain) into account. In this study, we aimed to address both of these issues by creating an incremental training approach for a torque prediction model using incomplete user-specific training data and biologically inspired temporal patterns of human gait. To reach this goal, we created a hybrid of two distinct approaches, a generic inter-individual and an adapting individual-specific model that exploits the inter-limb synergistic coupling during human gait to learn a function that predicts the torque at the ankle joint continuously based on the kinematic sequences of the hip, knee, and shank. An inter-individual generic base model learns temporal patterns of gait from a set of able-bodied individuals and predicts the gait patterns for a new individual, while the individual-specific adaptation model learns and predicts the temporal patterns of gait specific to a particular individual. The iterative training using the hybrid model was validated on eight able-bodied and five transtibial amputee subjects. It was found that, with the addition of estimators fitted to individual-specific data, the accuracy significantly increased from the baseline inter-individual model and plateaued within two to three iterations.

Task-specific Self-body Controller Acquisition by Musculoskeletal Humanoids: Application to Pedal Control in Autonomous Driving

Long-time Self-body Image Acquisition and its Application to the Control of Musculoskeletal Structures

Toward Autonomous Driving by Musculoskeletal Humanoids: A Study of Developed Hardware and Learning-Based Software

Learning Gait-conditioned Bipedal Locomotion with Motor Adaptation

Musculoskeletal AutoEncoder: A Unified Online Acquisition Method of Intersensory Networks for State Estimation, Control, and Simulation of Musculoskeletal Humanoids

Self-Body Image Acquisition and Posture Generation with Redundancy using Musculoskeletal Humanoid Shoulder Complex for Object Manipulation

Learning of Balance Controller Considering Changes in Body State for Musculoskeletal Humanoids

Online Self-body Image Acquisition Considering Changes in Muscle Routes Caused by Softness of Body Tissue for Tendon-driven Musculoskeletal Humanoids

Adaptive Whole-body Robotic Tool-use Learning on Low-rigidity Plastic-made Humanoids Using Vision and Tactile Sensors

Motion Modification Method of Musculoskeletal Humanoids by Human Teaching Using Muscle-Based Compensation Control

TWIMP: Two-Wheel Inverted Musculoskeletal Pendulum as a Learning Control Platform in the Real World with Environmental Physical Contact

An Exploratory Multi-Session Study of Learning High-Dimensional Body-Machine Interfacing for Assistive Robot Control

Learning-model-based Control for Robot Manipulators Sensorless Kinesthetic Teaching Using Sparse Feature Dynamics

Online Learning Feedback Control Considering Hysteresis for Musculoskeletal Structures

Self Model for Embodied Intelligence: Modeling Full-Body Human Musculoskeletal System and Locomotion Control with Hierarchical Low-Dimensional Representation

High Speed Whole Body Dynamic Motion Experiment with Real Time Master-Slave Humanoid Robot System

A hybrid approach for dynamically training a torque prediction model for devising a human-machine interface control strategy

Online Learning of Joint-Muscle Mapping Using Vision in Tendon-driven Musculoskeletal Humanoids

A Biological Humanoid Joint Controller Based on Muscle Model

Hierarchical Learning Framework for Whole-Body Model Predictive Control of a Real Humanoid Robot

Design and Control of a Pneumatic Musculoskeletal Biped Robot